Combining wavelength importance ranking to the random forest classifier to analyze multiclass spectral data

Area of Science:

• Analytical Chemistry
• Chemometrics
• Spectroscopy

Background:

• Near Infrared (NIR) spectroscopy is a powerful technique for substance characterization.
• NIR datasets often contain noise and correlated variables, which can hinder statistical analysis.
• Effective wavelength selection is crucial for optimizing classification performance in NIR spectroscopy.

Purpose of the Study:

• To evaluate the performance of the Random Forest (RF) classifier combined with wavelength importance ranking for classifying Near Infrared (NIR) spectral data.
• To develop and test a novel classification model integrating the chi-squared (χ²) ranking score with the RF classifier.
• To assess the model's effectiveness in reducing data dimensionality while improving classification accuracy across diverse product categories.

Main Methods:

• Utilized Random Forest (RF) classifier for sample classification.
• Implemented and compared various wavelength importance ranking approaches.
• Integrated the chi-squared (χ²) ranking score with the RF classifier for optimized wavelength selection.
• Validated the proposed method on six diverse NIR datasets (food, pharmaceuticals, illegal drugs).

Main Results:

• The proposed classification model significantly reduced the number of wavelengths required for analysis.
• Integration of the χ² ranking score and RF classifier led to increased classification accuracy compared to using complete NIR datasets.
• The method demonstrated robust performance across multiple datasets and classification tasks.
• Achieved superior results compared to existing competing methods in the literature.

Conclusions:

• The combination of χ² wavelength selection and RF classification offers an effective strategy for analyzing complex NIR spectral data.
• This approach enhances classification accuracy and reduces data dimensionality, making it suitable for practical applications.
• The findings suggest a valuable tool for quality control, authenticity verification, and substance identification using NIR spectroscopy.